25-Feb-2005
More than just a few chips…
Building Wireless Sensor Networks in 2005
Jan Beutel
© ETH Zürich | Computer Engineering and Networks Lab
225-Feb-2005 Jan Beutel – More than just a few chips…
Outline
Our device: BTnode rev3
Recent developments @ TIK BTnut System Software & Embedded
Development Deployment-Support Network TinyOS on the BTnode rev3
Experiences
325-Feb-2005 Jan Beutel – More than just a few chips…
New BTnode rev3: a lightweight dual radio
platform
BTnode success story Running both, TinyOS and the BTnut system software Prototyping Wireless Sensor Networks with BTnodes [EWSN2004]
3rd generation node commercialized with industrial partners AoT and Iftest Open-source policy has led to commercial replicas (Cobalt Blue by Vitronics)
58 mm
32.5 mm
425-Feb-2005 Jan Beutel – More than just a few chips…
It is very hard to deploy anywhere beyond 10-20 nodes.
Coordinated methods and tools are missing today.
Poor WSN development reality
525-Feb-2005 Jan Beutel – More than just a few chips…
BTnut Software & Embedded Development
BTnodes are not targeted at ultra low-power…
… but target versatile and flexible fast-prototyping.
Multi-threaded OS frame in C Standard open-source tools Lightweight software distribution
(8.2 MB binary, 27.3 MB source)
Rapid prototyping HW emulation on Linux PC
Demo applications and tutorial Lab for embedded systems lecture
simulate
emulate upload
compile
625-Feb-2005 Jan Beutel – More than just a few chips…
JAWS
ChallengeDevelopment and Deployment of Sensor Networks.
Application: Deployment-Support Networks
SolutionSelf-organizing backbone network with deployment-support services:
Virtual connections to nodes Remote reprogramming,
debugging, monitoring[SenSys2004], [IPSN2005]
traditional
725-Feb-2005 Jan Beutel – More than just a few chips…
Application: JAWS Prototype
825-Feb-2005 Jan Beutel – More than just a few chips…
Application: JAWS Experiments
Main obstacles Software complexity and stability Predictable operation and performance guarantees
Ongoing work JAWS permanent demo setup XTC routing implementation [Wattenhofer2001]
Distributed time synchronization [Blum2004]
Online runtime analysis BTnode power profiling [Negri – exchange PhD]
70+ node experiments
925-Feb-2005 Jan Beutel – More than just a few chips…
Related Work – Sensor Network DeploymentSimulation
TOSSIM [Levis2003 – Berkeley] PowerTOSSIM [Shnayder2004 – Harvard]
Testbeds Gnomes [Welsh2003 – Harvard] MoteLab [Wernerallen2005 – Harvard] Kansai – eXtreme Scaling Mote [OhioState] Mirage [Chun2005 – Intel]
Embedded emulation EmStar [Girod2004 – UCLA] BEE [Chang2003, Kuusilinna2003 – Berkeley]
Software distribution Deluge [Hui2004 - Berkeley]
1025-Feb-2005 Jan Beutel – More than just a few chips…
TinyOS on the BTnode rev3
Basics are working: BTnode rev3 is a Mica2 replacement Cooperation with Uni Copenhagen/ P. Bonnet BTnode3 platform definition available in contrib/tinybt Hard to “just port” without an application requirement because of
hardware dependencies in the software tinyos-1.x – 162 MB CVS nightmare nesC makes debugging harder, complexity is hidden within
Twin architecture of
Mica2 and BTnode 3
1125-Feb-2005 Jan Beutel – More than just a few chips…
Experiences 1 – where we stand todayHeterogeneous and large design space
No formal definition of a sensor network No single hardware/software platform for all/most needs Minimum required resource set for correct, reliable and
predictable function is unclear
Underestimated complexity Specialized chips are highly complex Applications need many support functions, software systems
grow
Keep it simple with the right level of abstraction Different people speak different languages Technology is only accessible to highly trained personnel, e.g.
not to an application-domain expert
1225-Feb-2005 Jan Beutel – More than just a few chips…
Experiences 2 – beyond the proof-of-conceptRealistic field testing is key to functional validation
Real-world interaction is necessary Sporadic race conditions do not show on a lab bench Standard embedded and real-time system issues are dominant In-situ testing, monitoring, calibration, validation and verification
have to be designed in
Emerging (open) standards? nesC 1.1 -> 1.2 major revision with new features will break
compatibility TinyOS 2.x will be completely new, looks promising, currently a
lot of “back to the roots” discussion in the 2.0 WG – component based – platform design – compatibility – reuse – many new platforms
1325-Feb-2005 Jan Beutel – More than just a few chips…
To probe further…
http://www.btnode.ethz.ch